Generalization in Mind and Machine

What do adversarial images tell us about human vision?

Autori: Marin Dujmović, Gaurav Malhotra, Jeffrey S Bowers
Pubblicato in: eLife, Numero 9, 2020, Pagina/e 1-29, ISSN 2050-084X
Editore: eLife Sciences Publications
DOI: 10.7554/elife.55978

Can deep convolutional neural networks support relational reasoning in the same-different task?

Autori: Guillermo Puebla; Jeffrey S. Bowers
Pubblicato in: Journal of Vision, Numero 22 (10) 11, 2022, Pagina/e 1-18, ISSN 1534-7362
Editore: Association for Research in Vision and Ophthalmology
DOI: 10.1167/jov.22.10.11

The role of capacity constraints in Convolutional Neural Networks for learning random versus natural data

Autori: Chris I Tsvetkov, Gaurav Malhotra, Benjamin D Evans, Jeffrey S Bowers
Pubblicato in: Neural Networks, Numero 161, 2023, Pagina/e 515-524, ISSN 0346-251X
Editore: Pergamon Press Ltd.
DOI: 10.1016/j.neunet.2023.01.011

Convolutional Neural Networks Are Not Invariant to Translation, but They Can Learn to Be

Autori: Valerio Biscione; Jeffrey S Bowers
Pubblicato in: Journal of Machine Learning Research, 2021, Pagina/e 1-28, ISSN 1532-4435
Editore: MIT Press
DOI: 10.48550/arxiv.2110.05861

Parallel Distributed Processing Theory in the Age of Deep Networks

Autori: Jeffrey S. Bowers
Pubblicato in: Trends in Cognitive Sciences, Numero 21/12, 2017, Pagina/e 950-961, ISSN 1364-6613
Editore: Elsevier BV
DOI: 10.1016/j.tics.2017.09.013

Feature Blindness: A Challenge for Understanding and Modelling Visual Object Recognition

Autori: Gaurav Malhotra; Marin Dujmovic; Jeffrey S Bowers
Pubblicato in: PLOS Computational Biology, 2022, ISSN 1932-6203
Editore: Public Library of Science
DOI: 10.1101/2021.10.20.465074

Biological convolutions improve DNN robustness to noise and generalisation

Autori: Benjamin D. Evans; Gaurav Malhotra; Jeffrey S. Bowers
Pubblicato in: Neural Networks, Numero 3, 2022, Pagina/e 96-110, ISSN 0893-6080
Editore: Pergamon Press Ltd.
DOI: 10.1016/j.neunet.2021.12.005

Successes and critical failures of neural networks in capturing human-like speech recognition

Autori: Federico Adolfi, Jeffrey S. Bowers, David Poeppel
Pubblicato in: Neural Networks, 2023, Pagina/e 199-211, ISSN 0893-6080
Editore: Pergamon Press Ltd.
DOI: 10.1016/j.neunet.2023.02.032

Training neural networks to encode symbols enables combinatorial generalization

Autori: Ivan I. Vankov, Jeffrey S. Bowers
Pubblicato in: Philosophical Transactions of the Royal Society B: Biological Sciences, Numero 375/1791, 2019, Pagina/e 20190309, ISSN 0962-8436
Editore: Royal Society of London
DOI: 10.1098/rstb.2019.0309

Hiding a plane with a pixel: examining shape-bias in CNNs and the benefit of building in biological constraints

Autori: Gaurav Malhotra, Benjamin D. Evans, Jeffrey S. Bowers
Pubblicato in: Vision Research, Numero 174, 2020, Pagina/e 57-68, ISSN 0042-6989
Editore: Pergamon Press Ltd.
DOI: 10.1016/j.visres.2020.04.013

Does that sound right? A Novel Method of Evaluating Models of Reading Aloud.

Autori: Michele Gubian, Ryan Blything, Colin Davis, Jeffrey Bowers
Pubblicato in: Behavior Research Methods, Numero 2022, 2022, ISSN 1554-351X
Editore: Springer Verlag
DOI: 10.3758/s13428-022-01794-8

Are there any ‘object detectors’ in the hidden layers of CNNs trained to identify objects or scenes?

Autori: Ella M. Gale, Nicholas Martin, Ryan Blything, Anh Nguyen, Jeffrey S. Bowers
Pubblicato in: Vision Research, Numero 176, 2020, Pagina/e 60-71, ISSN 0042-6989
Editore: Pergamon Press Ltd.
DOI: 10.1016/j.visres.2020.06.007

The human visual system and CNNs can both support robust online translation tolerance following extreme displacements

Autori: Ryan Blything, Valerio Biscione, Ivan I. Vankov, Casimir J. H. Ludwig, Jeffrey S. Bowers
Pubblicato in: Journal of Vision, Numero 21/2, 2021, Pagina/e 9, ISSN 1534-7362
Editore: Association for Research in Vision and Ophthalmology
DOI: 10.1167/jov.21.2.9

The Pitfalls of Measuring Representational Similarity Using Representational Similarity Analysis

Autori: Marin Dujmović, Jeffrey S. Bowers, Federico Adolfi, Gaurav Malhotra
Pubblicato in: BioRxiv, 2022
Editore: Cold Spring Harbor Laboratory
DOI: 10.1101/2022.04.05.487135

Generalisation in Neural Networks Does not Require Feature Overlap

Autori: Jeff Mitchell; Jeffrey S. Bowers
Pubblicato in: ArXiv, 2021
Editore: Cornell University
DOI: 10.48550/arxiv.2107.06872

Learning Online Visual Invariances for Novel Objects via Supervised and Self-Supervised Training

Autori: Valerio Biscione; Jeffrey S. Bowers
Pubblicato in: ArXiv, 2021
Editore: Cornell University
DOI: 10.48550/arxiv.2110.01476

Do DNNs Trained on Natural Images Organize Visual Features into Gestalts?

Autori: Valerio Biscione, Jeffrey S. Bowers
Pubblicato in: Arxiv, 2022
Editore: Arxiv
DOI: 10.48550/arxiv.2203.07302

Deep Problems with Neural Network Models of Human Vision

Autori: Jeffrey S. Bowers, Gaurav Malhotra, Marin Dujmović, Milton Llera Montero, Christian Tsvetkov, Biscione, Guillermo Puebla, Federico Adolfi, John E. Hummel, Rachel F. Heaton, Benjamin D. Evans, Jeffrey Mitchell, Ryan Blything
Pubblicato in: PsyArXiv, 2022
Editore: Cornell University
DOI: 10.31234/osf.io/5zf4s

Can deep convolutional neural networks support relational reasoning in the same-different task?

Autori: Guillermo Puebla; Jeffrey S. Bowers
Pubblicato in: BioRxiv, 2021
Editore: Cold Spring Harbor Laboratory
DOI: 10.1101/2021.09.03.458919

The contrasting shape representations that support object recognition in humans and CNNs

Autori: Gaurav Malhotra; Marin Dujmovic; John Hummel; Jeffrey S Bowers
Pubblicato in: BioRxiv, 2021
Editore: Cold Spring Harbor Laboratory
DOI: 10.1101/2021.12.14.472546

Lost in Latent Space: Disentangled Models and the Challenge of Combinatorial Generalisation

Autori: Milton Montero, Jeffrey S. Bowers, Rui Ponte Costa, Casimir J. Ludwig, Gaurav Malhotra
Pubblicato in: arXiv, 2022
Editore: Cornell Univesity
DOI: 10.48550/arxiv.2204.02283

A case for robust translation tolerance in humans and CNNs. A commentary on Han et al

Autori: Blything, Ryan; Biscione, Valerio; Bowers, Jeffrey
Pubblicato in: arXiv.org e-Print Archive, Numero 1, 2020
Editore: arXiv
DOI: 10.48550/arxiv.2012.05950

Learning Translation Invariance in CNNs

Autori: Valerio Biscione, Jeffrey Bowers
Pubblicato in: NeurIPS 2020 Workshop SVRHM, 2020
Editore: NeurIPS 2020 Workshop SVRHM

The role of disentanglement in generalisation

Autori: Milton Llera Montero; Casimir Ludwig; Rui Ponte Costa; Gaurav Malhotra; Jeffrey S Bowers
Pubblicato in: International Conference on Learning Representations, 2021
Editore: OpenReview

Can Deep Convolutional Neural Networks Learn Same-Different Relations?

Autori: Guillermo Puebla; Jeffrey S. Bowers
Pubblicato in: Proceedings of the Annual Meeting of the Cognitive Science Society, 43(43), Numero 43(43), 2021, ISSN 1069-7977
Editore: University of California

Subtractive gating improves generalization in working memory tasks

Autori: Milton Llera Montero, Gaurav Malhotra, Jeff Bowers, Rui Ponte Costa
Pubblicato in: 2019 Conference on Cognitive Computational Neuroscience, 2019
Editore: Cognitive Computational Neuroscience
DOI: 10.32470/ccn.2019.1352-0

Do LSTMs know about Principle C?

Autori: Jeff Mitchell, Nina Kazanina, Conor Houghton, Jeff Bowers
Pubblicato in: 2019 Conference on Cognitive Computational Neuroscience, 2019
Editore: Cognitive Computational Neuroscience
DOI: 10.32470/ccn.2019.1241-0

Extreme Translation Tolerance in Humans and Machines

Autori: Ryan Blything, Ivan Vankov, Casimir Ludwig, Jeffrey Bowers
Pubblicato in: 2019 Conference on Cognitive Computational Neuroscience, 2019
Editore: Cognitive Computational Neuroscience
DOI: 10.32470/ccn.2019.1091-0

Humans cannot decipher adversarial images: Revisiting Zhou and Firestone (2019)

Autori: Marin Dujmović, Gaurav Malhotra, Jeffrey Bowers
Pubblicato in: 2019 Conference on Cognitive Computational Neuroscience, 2019
Editore: Cognitive Computational Neuroscience
DOI: 10.32470/ccn.2019.1298-0

Adding biological constraints to CNNs makes image classification more human-like and robust

Autori: Gaurav Malhotra, Benjamin Evans, Jeffrey Bowers
Pubblicato in: 2019 Conference on Cognitive Computational Neuroscience, 2019
Editore: Cognitive Computational Neuroscience
DOI: 10.32470/ccn.2019.1212-0

Translation Tolerance in Vision.

Autori: Ryan Blything, Ivan I. Vankov, Casimir J. Ludwig, Jeffrey S. Bowers
Pubblicato in: 41st Annual Conference of the Cognitive Science Society 2019, 2019
Editore: A.K. Goel, C.M. Seifert, & C. Freksa (Eds.)

Selectivity Metrics Provide Misleading Estimates of the Selectivity of Single Units in Neural Networks

Autori: Ella Gale, Ryan Blything, Nicholas Martin, Jeffrey S. Bowers, and Anh Nguyen.
Pubblicato in: 41st Annual Conference of the Cognitive Science Society 2019, 2019
Editore: A.K. Goel, C.M. Seifert, & C. Freksa (Eds.)

The Contrasting Roles of Shape in Human Vision and Convolutional Neural Networks

Autori: Gaurav Malhotra, Jeff Bowers
Pubblicato in: 41st Annual Conference of the Cognitive Science Society 2019, 2019
Editore: A.K. Goel, C.M. Seifert, & C. Freksa (Eds.)

Harnessing the Symmetry of Convolutions for Systematic Generalisation

Autori: Jeff Mitchell, Jeffrey S. Bowers
Pubblicato in: 2020 International Joint Conference on Neural Networks (IJCNN), 2020, Pagina/e 1-8, ISBN 978-1-7281-6926-2
Editore: IEEE
DOI: 10.1109/ijcnn48605.2020.9207183

Adding biological constraints to deep neural networks reduces their capacity to learn unstructured data

Autori: Christian Tsvetkov, Gaurav Malhotra, Benjamin Evans, Jeff Bowers
Pubblicato in: In Proceedings of the 42nd Annual Conference of the Cognitive Science Society 2020, 2020, Pagina/e 2358-2364
Editore: Cognitive Science Society

Priorless Recurrent Networks Learn Curiously

Autori: Jeff Mitchell, Jeffrey Bowers
Pubblicato in: Proceedings of the 28th International Conference on Computational Linguistics, 2020, Pagina/e 5147-5158, ISBN 978-1-952148-27-9
Editore: International Committee on Computational Linguistics
DOI: 10.18653/v1/2020.coling-main.451

Pubblicazioni

Condividi questa pagina

Scarica